Dae‐Shik Kim

Seonghak Kim, Gyeongdo Ham, Suin Lee et al.

To apply the latest computer vision techniques that require a large computational cost in real industrial applications, knowledge distillation methods (KDs) are essential. Existing logit-based KDs apply the constant temperature scaling to all samples in dataset, limiting the utilization of knowledge inherent in each sample individually. In our approach, we classify the dataset into two categories (i.e., low energy and high energy samples) based on their energy score. Through experiments, we have confirmed that low energy samples exhibit high confidence scores, indicating certain predictions, while high energy samples yield low confidence scores, meaning uncertain predictions. To distill optimal knowledge by adjusting non-target class predictions, we apply a higher temperature to low energy samples to create smoother distributions and a lower temperature to high energy samples to achieve sharper distributions. When compared to previous logit-based and feature-based methods, our energy-based KD (Energy KD) achieves better performance on various datasets. Especially, Energy KD shows significant improvements on CIFAR-100-LT and ImageNet datasets, which contain many challenging samples. Furthermore, we propose high energy-based data augmentation (HE-DA) for further improving the performance. We demonstrate that higher performance improvement could be achieved by augmenting only a portion of the dataset rather than the entire dataset, suggesting that it can be employed on resource-limited devices. To the best of our knowledge, this paper represents the first attempt to make use of energy function in knowledge distillation and data augmentation, and we believe it will greatly contribute to future research.

Gyeongdo Ham, Seonghak Kim, Suin Lee et al.

Previous logits-based Knowledge Distillation (KD) have utilized predictions about multiple categories within each sample (i.e., class predictions) and have employed Kullback-Leibler (KL) divergence to reduce the discrepancy between the student and teacher predictions. Despite the proliferation of KD techniques, the student model continues to fall short of achieving a similar level as teachers. In response, we introduce a novel and effective KD method capable of achieving results on par with or superior to the teacher models performance. We utilize teacher and student predictions about multiple samples for each category (i.e., batch predictions) and apply cosine similarity, a commonly used technique in Natural Language Processing (NLP) for measuring the resemblance between text embeddings. This metric's inherent scale-invariance property, which relies solely on vector direction and not magnitude, allows the student to dynamically learn from the teacher's knowledge, rather than being bound by a fixed distribution of the teacher's knowledge. Furthermore, we propose a method called cosine similarity weighted temperature (CSWT) to improve the performance. CSWT reduces the temperature scaling in KD when the cosine similarity between the student and teacher models is high, and conversely, it increases the temperature scaling when the cosine similarity is low. This adjustment optimizes the transfer of information from the teacher to the student model. Extensive experimental results show that our proposed method serves as a viable alternative to existing methods. We anticipate that this approach will offer valuable insights for future research on model compression.

Suin Lee, Dae-Shik Kim

We present TexTailor, a novel method for generating consistent object textures from textual descriptions. Existing text-to-texture synthesis approaches utilize depth-aware diffusion models to progressively generate images and synthesize textures across predefined multiple viewpoints. However, these approaches lead to a gradual shift in texture properties across viewpoints due to (1) insufficient integration of previously synthesized textures at each viewpoint during the diffusion process and (2) the autoregressive nature of the texture synthesis process. Moreover, the predefined selection of camera positions, which does not account for the object's geometry, limits the effective use of texture information synthesized from different viewpoints, ultimately degrading overall texture consistency. In TexTailor, we address these issues by (1) applying a resampling scheme that repeatedly integrates information from previously synthesized textures within the diffusion process, and (2) fine-tuning a depth-aware diffusion model on these resampled textures. During this process, we observed that using only a few training images restricts the model's original ability to generate high-fidelity images aligned with the conditioning, and therefore propose an performance preservation loss to mitigate this issue. Additionally, we improve the synthesis of view-consistent textures by adaptively adjusting camera positions based on the object's geometry. Experiments on a subset of the Objaverse dataset and the ShapeNet car dataset demonstrate that TexTailor outperforms state-of-the-art methods in synthesizing view-consistent textures. The source code for TexTailor is available at https://github.com/Adios42/Textailor

Vyacheslav Shen, Kassymzhomart Kunanbayev, Dae-Shik Kim

With the advancements in Large Language and Latent Diffusion models, brain decoding has achieved remarkable results in recent years. The works on the NSD dataset, with stimuli images from the COCO dataset, leverage the embeddings from the CLIP model for image reconstruction and GIT for captioning. However, the current captioning approach introduces the challenge of potential data contamination given that the GIT model was trained on the COCO dataset. In this work, we present an alternative method for decoding brain signals into image captions by predicting a DINOv2 model's embedding of an image from the corresponding fMRI signal and then providing its [CLS] token as the prefix to the GPT-2 language model which decreases computational requirements considerably. Additionally, instead of commonly used Linear Regression, we explore 3D Convolutional Neural Network mapping of fMRI signals to image embedding space for better accounting positional information of voxels.

Nahyun Kim, Donggon Jang, Sunhyeok Lee et al.

Supervised learning-based methods yield robust denoising results, yet they are inherently limited by the need for large-scale clean/noisy paired datasets. The use of unsupervised denoisers, on the other hand, necessitates a more detailed understanding of the underlying image statistics. In particular, it is well known that apparent differences between clean and noisy images are most prominent on high-frequency bands, justifying the use of low-pass filters as part of conventional image preprocessing steps. However, most learning-based denoising methods utilize only one-sided information from the spatial domain without considering frequency domain information. To address this limitation, in this study we propose a frequency-sensitive unsupervised denoising method. To this end, a generative adversarial network (GAN) is used as a base structure. Subsequently, we include spectral discriminator and frequency reconstruction loss to transfer frequency knowledge into the generator. Results using natural and synthetic datasets indicate that our unsupervised learning method augmented with frequency information achieves state-of-the-art denoising performance, suggesting that frequency domain information could be a viable factor in improving the overall performance of unsupervised learning-based methods.

Goutam Bhat, Martin Danelljan, Radu Timofte et al.

This paper reviews the NTIRE2021 challenge on burst super-resolution. Given a RAW noisy burst as input, the task in the challenge was to generate a clean RGB image with 4 times higher resolution. The challenge contained two tracks; Track 1 evaluating on synthetically generated data, and Track 2 using real-world bursts from mobile camera. In the final testing phase, 6 teams submitted results using a diverse set of solutions. The top-performing methods set a new state-of-the-art for the burst super-resolution task.

Andriy Serdega, Dae-Shik Kim

Learning interpretable and disentangled representations of data is a key topic in machine learning research. Variational Autoencoder (VAE) is a scalable method for learning directed latent variable models of complex data. It employs a clear and interpretable objective that can be easily optimized. However, this objective does not provide an explicit measure for the quality of latent variable representations which may result in their poor quality. We propose Variational Mutual Information Maximization Framework for VAE to address this issue. In comparison to other methods, it provides an explicit objective that maximizes lower bound on mutual information between latent codes and observations. The objective acts as a regularizer that forces VAE to not ignore the latent variable and allows one to select particular components of it to be most informative with respect to the observations. On top of that, the proposed framework provides a way to evaluate mutual information between latent codes and observations for a fixed VAE model. We have conducted our experiments on VAE models with Gaussian and joint Gaussian and discrete latent variables. Our results illustrate that the proposed approach strengthens relationships between latent codes and observations and improves learned representations.

Andriy Serdega, Dae-Shik Kim

Variational Autoencoder is a scalable method for learning latent variable models of complex data. It employs a clear objective that can be easily optimized. However, it does not explicitly measure the quality of learned representations. We propose a Variational Mutual Information Maximization Framework for VAE to address this issue. It provides an objective that maximizes the mutual information between latent codes and observations. The objective acts as a regularizer that forces VAE to not ignore the latent code and allows one to select particular components of it to be most informative with respect to the observations. On top of that, the proposed framework provides a way to evaluate mutual information between latent codes and observations for a fixed VAE model.

Geondo Park, Chihye Han, Wonjun Yoon et al.

Visual-semantic embedding enables various tasks such as image-text retrieval, image captioning, and visual question answering. The key to successful visual-semantic embedding is to express visual and textual data properly by accounting for their intricate relationship. While previous studies have achieved much advance by encoding the visual and textual data into a joint space where similar concepts are closely located, they often represent data by a single vector ignoring the presence of multiple important components in an image or text. Thus, in addition to the joint embedding space, we propose a novel multi-head self-attention network to capture various components of visual and textual data by attending to important parts in data. Our approach achieves the new state-of-the-art results in image-text retrieval tasks on MS-COCO and Flicker30K datasets. Through the visualization of the attention maps that capture distinct semantic components at multiple positions in the image and the text, we demonstrate that our method achieves an effective and interpretable visual-semantic joint space.

Deokyun Kim, Minseon Kim, Gihyun Kwon et al.

Face Super-Resolution (SR) is a subfield of the SR domain that specifically targets the reconstruction of face images. The main challenge of face SR is to restore essential facial features without distortion. We propose a novel face SR method that generates photo-realistic 8x super-resolved face images with fully retained facial details. To that end, we adopt a progressive training method, which allows stable training by splitting the network into successive steps, each producing output with a progressively higher resolution. We also propose a novel facial attention loss and apply it at each step to focus on restoring facial attributes in greater details by multiplying the pixel difference and heatmap values. Lastly, we propose a compressed version of the state-of-the-art face alignment network (FAN) for landmark heatmap extraction. With the proposed FAN, we can extract the heatmaps suitable for face SR and also reduce the overall training time. Experimental results verify that our method outperforms state-of-the-art methods in both qualitative and quantitative measurements, especially in perceptual quality.

Gihyun Kwon, Chihye Han, Dae-shik Kim

As deep learning is showing unprecedented success in medical image analysis tasks, the lack of sufficient medical data is emerging as a critical problem. While recent attempts to solve the limited data problem using Generative Adversarial Networks (GAN) have been successful in generating realistic images with diversity, most of them are based on image-to-image translation and thus require extensive datasets from different domains. Here, we propose a novel model that can successfully generate 3D brain MRI data from random vectors by learning the data distribution. Our 3D GAN model solves both image blurriness and mode collapse problems by leveraging alpha-GAN that combines the advantages of Variational Auto-Encoder (VAE) and GAN with an additional code discriminator network. We also use the Wasserstein GAN with Gradient Penalty (WGAN-GP) loss to lower the training instability. To demonstrate the effectiveness of our model, we generate new images of normal brain MRI and show that our model outperforms baseline models in both quantitative and qualitative measurements. We also train the model to synthesize brain disorder MRI data to demonstrate the wide applicability of our model. Our results suggest that the proposed model can successfully generate various types and modalities of 3D whole brain volumes from a small set of training data.